AI Company Anthropic has called on major artificial intelligence labs to agree on a coordinated pause in advanced AI development.
The company warned that AI systems are improving at a very fast pace, which could increase risks for society if not properly controlled. Anthropic said that the ability of AI systems to complete tasks on their own is doubling about every four months. It warned that the industry could soon reach a stage where AI can improve itself without human help. This stage is often called recursive self-improvement.
The company said that if AI systems start building better versions of themselves, safety and control will become much more important. It added that society needs time to understand and manage the risks before reaching that point.
Anthropic co-founders Jack Clark and Marina Favaro said such a moment may come sooner than many institutions expect. They also said that it is not guaranteed, but it is possible in the near future.
Concerns about powerful AI systems have increased as the technology becomes more advanced. Some experts fear that highly capable AI could become difficult to control and may create risks for jobs, security, and society. Earlier this year, Anthropic’s AI model showed strong performance in detecting weaknesses in software systems. This raised concerns in industries like banking and software security.
At the same time, global regulation has not kept up with the fast growth of AI. In the United States, where many major AI companies are based, most rules are still voluntary. A recent executive order under the Trump administration encouraged AI companies to share their most advanced models for government testing before public release. However, this is still based on cooperation rather than strict law.
Previous calls for a pause in AI development have been made by researchers and tech leaders, including Elon Musk and others. A similar proposal in 2023 suggested stopping advanced AI work for six months to create safety rules, but it was not widely adopted.
Anthropic says it remains focused on AI safety. Earlier, it restricted some uses of its models, including military surveillance and autonomous weapons. This led to tension with US authorities, although reports now suggest relations may be improving.
Our internal data shows Claude is accelerating AI development—a possible path to recursive self-improvement, or AI autonomously building a more capable successor.
It’s happening faster than we thought, and the implications deserve greater attention. https://t.co/OVVPJO7VQx
— Anthropic (@AnthropicAI) June 4, 2026
The company has also continued building more powerful AI systems. In recent months, it adjusted some of its earlier safety limits, saying it would not hold back development if competitors move faster. Anthropic is now valued at a very high level after major investment funding. It has also filed for a possible public share listing in the United States.
The company said that a meaningful pause would need agreement between several leading AI labs. It added that rules would also be needed to decide when such a pause should start and end. Anthropic warned that if only one company slows down, others may continue developing faster. It said this would not solve the wider saftey problem and could even shift competition rather than reduce risk.
The company plans to bring together researchers, policymakers, and industry leaders in the coming months. The goal is to discuss how to manage advanced AI systems and reduce possible future risks.
